Oxygen Minimum Zones (Omzs) in the Modern Ocean

Progress in Oceanography 80 (2009) 113–128

Contents lists available at ScienceDirect

Progress in Oceanography

journal homepage: www.elsevier.com/locate/pocean

Oxygen minimum zones (OMZs) in the modern ocean

A. Paulmier a,b,*, D. Ruiz-Pino b a LEGOS/CNRS 18, Av. Ed. Belin, 31401 Toulouse Cedex 9, France b LOCEAN, Université P&M Curie, Courrier 134, 4 pl. Jussieu, 75252 Paris Cedex 05, France article info abstract

Article history: In the modern ocean, oxygen minimum zones (OMZs) are potential traces of a primitive ocean in Received 6 September 2007 which Archean bacteria lived and reduced chemical anomalies occurred. But OMZs are also keys to Received in revised form 1 August 2008 understanding the present unbalanced nitrogen cycle and the oceans’ role on atmospheric greenhouse Accepted 4 August 2008 control. OMZs are the main areas of nitrogen loss (as N ,NO) to the atmosphere through denitriﬁ- Available online 17 August 2008 2 2 cation and anammox, and could even indirectly mitigate the oceanic biological sequestration of

CO2. It was recently hypothesized that OMZs are going to spread in the coming decades as a conse- Keywords: quence of global climate change. Despite an important OMZ role for the origin of marine life and for Oxygen minimum zones (OMZs) the biogeochemical cycles of carbon and nitrogen, there are some key questions on the structure of Oxygen Global ocean OMZs at a global scale. There is no agreement concerning the threshold in oxygen that defines an Denitrification OMZ, and the extent of an OMZ is often evaluated by denitrification criteria which, at the same time, Biogeochemistry are O2-dependent. Our work deals with the identification of each OMZ, the evaluation of its extent, volume and vertical structure, the determination of its seasonality or permanence and the comparison between OMZs and denitrification zones at a global scale. The co-existence in the OMZ of oxic (in its boundaries) and suboxic (even anoxic, in its core) conditions involves rather complex biogeochemical processes such as strong remineralization of the organic matter, removal of nitrate and release of nitrite. The quan- titative OMZ analysis is focused on taking into account the whole water volume under the influence of an OMZ and adapted to the study of the specific low oxygen biogeochemical processes.

A characterization of the entire structure for the main and most intense OMZs (O2 <20lM reaching 1 lM in the core) is proposed based on a previously published CRIO criterion from the eastern South Pacific OMZ and including a large range of O2 concentrations. Using the updated global WOA2005 O2 climatology, the four known tropical OMZs in the open ocean have been described: the Eastern South Pacific and Eastern Tropical North Pacific, in the Pacific Ocean; the Arabian Sea and Bay of Bengal, in the Indian Ocean. Moreover, the Eastern Sub-Tropical North Pacific (25–52°N) has been identified as a lesser known permanent deep OMZ. Two additional seasonal OMZs at high latitude have also been identified: the West Bering Sea and the Gulf of Alaska. The total surface of the permanent OMZs is 30.4 millions of km2 ( 8% of the total oceanic area), and the volume of the OMZ cores (10.3 millions of km3) corresponds to a value 7 times higher than previous evaluations. The volume of the OMZ cores is about three times larger than that of the associated denitrification zone, here defined as NMZ (‘nitrate deficit or NDEF > 10 lM’ maximum zone). The larger OMZ, relative to the extent of denitrification zone, suggests that the unbalanced nitrogen cycle on the global scale could be more intense than previously recognized and that evaluation of the OMZ from denitrification could underestimate their extent. Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction (1972). OMZs correspond to subsurface oceanic zones (e.g., at 50– 100 m depth in the Arabian Sea; Morrison et al., 1999) and reaching

The interest in oxygen minimum zones (OMZs), characterized ultra-low values of O2 concentration (e.g. <1 lM; Karstensen et al., herein as O2-deﬁcient layers in the ocean water column, is quite re- 2008). OMZs, because of their intensity and shallowness, are, a priori, cent, since the appearance of the name ‘‘OMZ” in Cline and Richards different from the relatively well known ‘‘classical O2 minimum”, which is 50 times more oxygenated than OMZs and found at intermediate depths (1000–1500 m) in all the oceans (Wyrtki, * Corresponding author. Address: LEGOS/CNRS 18, Av. Ed. Belin, 31401 Toulouse Cedex 9, France. Tel.: +33 (0)561333007; fax: +33 (0)561253205. 1962). Note that in the present study, an OMZ is deﬁned as being E-mail address: [email protected] (D. Ruiz-Pino). ‘‘more intense’’, when the O2 concentrations in its core are lower.

1.1. OMZ specificities for marine biogeochemistry and ecosystems conditions, OMZs would increase or intensify, according to observations in recent decades (e.g., Stramma et al., 2008). But evalua- OMZs have been mainly known for playing an essential role in tions or predictions of OMZs variation over paleoclimatic periods, the global nitrogen cycle, in which various chemical species, since the anthropocene era or in the future, cannot be validated according to their degree of oxidation (e.g. ammonium, NH4þ; without a reference state, and the report of all the existing OMZs nitrite, NO2À; nitrate, NO3À; nitrous oxide, N2O; dinitrogen, N2), detected in the modern ocean taking into account improvements and different bacterial processes intervene. Under oxic conditions, in O2-measurement techniques. but also at the upper boundary (oxycline) of an OMZ, nitrification Despite the important role of OMZs in understanding primitive transforms NH4þ into NO3À. But OMZs are especially associated with marine life and chemistry, as well as in the carbon (C) and nitrogen denitrification, which is a bacterial process occurring only in (N) cycles, little knowledge has been obtained on the extent and

O2-deficient regions (e.g., Codispoti et al., 2001). Denitrification vertical structure of these oceanic ‘‘curiosities”. This is mainly converts NO3À, one of the main limiting nutrients in the ocean, into due to the following difficulties: (i) few available O2 data obtained gaseous nitrogen (N as, for example, N2O, N2) which is lost to the with a low enough detection limit (<1 lM) and accuracy (<2 lM), atmosphere and contributes to the oceanic nitrate deficit (N/ owing to the present limitations in the sampling and analysis tech- P 14.7; e.g., Tyrrell, 1999). However, recently, an unknown pro- niques linked to the low O concentration; (ii) the choice of a 2 cess in the ocean has been observed, first in sediments and then unique criterion for all OMZs, since the nature of this criterion in the water column in the OMZs (e.g., Kuypers et al., 2003): the often depends on research interest (e.g., specific low-O2 biogeo- anaerobic oxidation of NH4þ using NO2À (anammox); this imposes chemistry process studies have to take into account an O2 concen- a complete revision of the global nitrogen cycle (e.g., Arrigo, tration lower than 20 lM, but the influence of physical processes 2005). OMZs are also involved in the cycle of very important cli- do not make it necessary to include suboxic and anoxic condi- matic gases: (i) production of 50% of the oceanic N O (e.g., Bange tions); (iii) the criteria could be different for each OMZ region: 2 et al., 1996); (ii) production of H2 S (e.g., Dugdale et al., 1977) and for example, the OMZs in the Northeastern Atlantic ocean is ex- CH4 (e.g., Cicerone and Oremland, 1988), episodically or for OMZs cluded when a threshold of 20 lM is used (Helly and Levin, in contact with sediments; (iii) limitation of atmospheric CO2 2004). Different terms and thresholds have been used to described sequestration by the ocean: directly as an end-product of reminer- the overall low O2 conditions. Suboxia has been mainly defined by alization (Paulmier et al., 2006) or indirectly through limitation of biologists and biogeochemists as a transition layer from O2-to total primary production due to the N loss (see hypothesis of Fal- NOÀ-respiration, with thresholds between 0.7 lM (e.g., Yakusev 3 kowski, 1997); (iv) potential DMS consumption due to higher bac- and Neretin, 1997) and 20 lM (e.g., Helly and Levin, 2004). Hypox- terial activity (Kiene and Bates, 1990). Chemically, OMZs are ia implies O2 conditions under which macro-organisms cannot live: associated with acidification (low pH 7.5 SWS; Paulmier, 2005), 8 lM for Kamykowski and Zentara (1990), but up to 40 lM and reduced conditions (Richards, 1965) favoring reduced chemi- depending on the species considered, such as anchovy (e.g., Gray cal species (e.g., Fe(II) or Cu(I) potentially stimulating photosynthe- et al., 2002). Dysoxia (O2 <4lM) and microxia (O2 <1lM; Levin, sis or N2O production). 2002) are associated with a sharp O2 transition for the large organ- OMZs have also increased interest in biological and ecosystem isms, such as fishes. Anoxia (O2 < 0.1 lM; Oguz et al., 2000) is defined studies. Because of similarities between Archean bacteria and those by transition from NO3-respiration to sulphate-reduction. living in the OMZs (Zumft, 1997), OMZs could be considered as ana- The first global study providing information on where water logues of the primitive anoxic ocean in which life is widely thought column OMZs can be located is that of Kamykowski and Zentara to have first appeared. Transitions from high to low (the appearance (1990) who produced maps of the distribution of hypoxia (O2 < of OMZs) oxygenation periods could stimulate biodiversity on a 8 lM) and of denitrification (Nitrate DEFicit or NDEF > 10 lM): paleoclimatic scale (Rogers, 2000). OMZs can be a refuge for organ- ENP (Eastern North Pacific), ESP (Eastern South Pacific), AS (Arabian isms specifically adapted to low O2 concentration (e.g., giant Thiop- Sea) and BB (Bay of Bengal; see Fig. 1a). Without having a known loca bacteria; Levin, 2002) from predation or competition with evaluation of an OMZ’s surface and vertical structure, Codispoti other species, and the lower OMZ boundary can even be among et al. (2001) concluded that the volume of suboxic zones could the richest habitats for the megafauna of the ocean. As a respiratory reach 0.1% of the oceanic volume. OMZ areas have been consid- barrier, OMZs are associated with active vertical daily migration ered to be similar to those of denitrification in several regional (e.g., for zooplankton; Fernández-Alamo and Färber-Lorda, 2006). studies: ENP, ESP, AS (e.g., Codispoti et al., 2001). Hattori (1983) But, for the main species (e.g., commercial fishes, such as anchovy), evaluated global oceanic denitrification ( 8.45 106 km2), Â OMZs are considered as inhospitable. In the past, the Oceanic An- obtained from separate previous evaluations using different criteria oxic Events (OAEs) have been associated with massive species for each OMZ: NDEF > 10 lM (AS); secondary subsurface NO2À peak extinction (e.g. during the Mid-Cretaceous). In the present, episodic (ESP); O2 <5lM (ENP). But from a qualitative comparison of hy- anoxic events associated with eutrophicated waters are also induc- poxia and denitrification maps, Kamykowski and Zentara (1990) ing massive abnormal fish mortality (e.g., Chan et al., 2008). concluded that the extent of the denitrification zone would be much less than the extent of the OMZ. Such a difference shows that

1.2. Need for a reference state and an O2 criterion for defining the the denitrification criterion could not be adapted to evaluating the OMZs extent size of the whole O2-deficit zone. To validate this hypothesis, it is necessary to evaluate OMZs independently of the extent of the The intensity of all OMZ’s perturbations and their potential denitrification zone. Note also that the surface of OMZs that is in feedback to climate and the marine ecosystem would depend on contact with sediments (1372 106 km2; Helly and Levin, 2004) Â the OMZs extent. This extent would vary in response to climatic would be an order of magnitude lower than the global denitrifica- changes (lower ventilation due to stratification, and decreased O2 tion zone and the associated water-column OMZ surfaces. solubility) and natural or anthropogenic fertilization (increased Estimations of the extent of OMZs for biogeochemical studies remineralization) through nutrient or metal inputs by upwelling, are scarce and/or local (e.g., Morrison et al., 1999). Recently, the river discharge or atmospheric dust fall-out (e.g., Béthoux, 1989; quantification of OMZs in the open ocean has been proposed by

Joos et al., 2003). In the past, OMZs have probably extended and Karstensen et al. (2008). In the present study, three O2 thresholds contracted in warm (interglacial) and cold (glacial) periods, respec- were used (the suboxic level of 4.5 lmol/kg, a more stringent tively (e.g., Cannariato and Kennett, 1999). Under present-day 45 lmol/kg and a more relaxed level of 90 lmol/kg), and the anal- A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 115

Fig. 1. (a) O2 distribution (lM) at depth where O2 concentration is minimal, indicating the extent of the OMZs (in red) according to the WOA2005 climatology. The color bar scale corresponds to a 1 ± 2 lM interval between 0 and 20 lM, and a 20 ± 2 lM interval between 20 and 340 lM. The isolines indicate the limit of the upper OMZ CORE depth in meters with a 100-m contour interval. For OMZ acronyms, see the list at the end of the main text. (b) NDEF > 10 ± 2.5 lM distribution (lM) at depth where NDEF is maximal, indicating the extent of the NMZs from the WOA2005 climatology. The extent of the OMZs (see (a), above) is marked by contours in black.

yses were focused on the tropical Atlantic and Pacific Oceans. The evaluation by Morrison et al. (1999). Why another evaluation of OMZ volumes thus evaluated were of 0.461, 18.6 and 38.3 OMZ extent and volume? Because our focus is on defining an 6 3 Â 10 km for each proposed O2 threshold, respectively. Although OMZ structure and extent which allows us to take into account spe- not focused on the OMZs in the Indian Ocean, Karstensen et al. cific biogeochemical processes, such as denitrification or anammox,

(2008) proposed an estimation of the vertical extent for the OMZ associated with low O2. The evaluation and criteria proposed by in the Arabian Sea of 550 m, i.e. about twice as small as the local Karstensen and coauthors are more adapted to the analysis of the 116 A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 dynamical processes responsible for the formation of an OMZ, threshold of 20 lM could be used with sufficient confidence, based though excluding the formation of OMZs in the Indian Ocean, where on the O2 detection limit and the uncertainties ( 20 lM) of the probably the most intense denitrification and nitrogen loss occur, main O2 databases available. Using O2 <20 lM, the CRIO criterion and such as we will see here, do not include OMZs at a high subtrop- excludes the OMZs (or low O2 zones called, LOZ) in the open tropical latitude. It was shown from an analysis of the ESP OMZ off Chile ical Atlantic Ocean (O2 > 40, and 20 lM in the Canary and Benguela (Paulmier et al., 2006) that the existence of three different layers Current systems, respectively; Karstensen et al., 2008), in which no has to be taken into account to evaluate the entire OMZ structure: denitrification has yet been reported, except on the continental the oxycline (upper O gradient, 5 times more intense than in the margin (e.g., in the Benguela Current system). The present study 2 oxygenated ocean); the core (O2 <20lM); the lower O2 gradient. therefore focuses on the most intense OMZs of the open ocean, Indeed, the oxycline is considered as the OMZ engine, where the reaching the weakest concentrations (down to O2 <1 lM) in the most intense remineralization occurs, leading to the OMZ’s intensi- eastern Pacific Ocean and the northern Indian Ocean. In addition fication, and where a specific denitrification and nitrification cou- to the CORE, the upper OMZ boundary layer border, called the oxy- pling (e.g., Brandes et al., 2007) could also occur with O2 >20lM. cline (OXY), which plays a role as an OMZ biogeochemical engine OMZ core, specific to anaerobic processes as canonical (classical (Paulmier et al., 2006), is defined by gradients higher than anaerobic) denitrification, and the lower O2 gradient, where nitrifi- 0.9 lM/m, as for the OMZ off Chile. The lower OMZ boundary layer, cation is a main process, could play an important role in the nitro- called the lower O2 gradient (LOG), is delimited by the depth at gen cycling in the OMZ (e.g., Anderson et al., 1982). Thus, to which the O2 gradient becomes less than 0.1 lM/m, corresponding consider the specific biogeochemical processes, it is necessary to in- to the strongest O2 gradient for the ‘‘classical O2 minimum”. clude these three layers and the large range of O2 concentrations, and not only the extremely low O2 observed in the OMZ core. 2.2. Denitrification criteria for NMZ (‘‘nitrate deficit” maximum zone) Finally, having in mind to answer the question of how denitrification criteria are or are not adapted to the evaluation of the extent Previous indirect quantifications of the vertical and horizontal of an OMZ, it is necessary to determine simultaneously the struc- extents of an OMZ used a criterion based on the denitrification ture and extent of OMZ and the denitrification zones. activity, which focuses mainly on the calculation of different indi-

Hence, from the same O2 criterion and comparison with the cri- ces (e.g., Hattori, 1983): the NO3À deficit (NDEF > 10 lM) and/or the teria for denitrification, the main and most intense OMZs in the NO2À secondary subsurface peak (>5 lM). open ocean are identified and characterized quantitatively (hori- Denitrification was evaluated quantitatively with NDEF ap- zontally and vertically). The permanence and potential seasonality proach and compared qualitatively with the subsurface NO2À peak, of the OMZs will be analyzed. However, OMZs formed over the which also indicates the presence of denitrification (NO3À-reduction continental shelf (such as the Benguela OMZ) and in semi-enclosed into NO2À; Codispoti and Christensen, 1985). The ‘NDEF > 10 lM’ seas (such as the Black Sea) or over deep trenches (e.g. Gulf of Car- criterion corresponding to the historical definition (NDEF 3 ¼ iaco, Venezuela) reaching the level of anoxia will not be addressed 15PO À NOÀ; Broecker and Peng, 1982) and previously used at 4 À 3 in this study. the global scale (Kamykowski and Zentara, 1990) will be here determined. N* (Gruber and Sarmiento, 1997) was not chosen, because the threshold corresponding to significant denitrification has not 2. Methodology yet been well defined, absolute N* values being arbitrary (Gruber, 2004), although the same conclusions as with NDEF can be obtained To characterize and determine the surface and volume of an with N* < 9 lM. Thus here, from the computation of NDEF, and by À OMZ, the CRIO criterion on O2, adapted to take into account the en- analogy with the OMZ, an NMZ (NDEF maximum zone) has been tire vertical thickness of an OMZ, and compared with the denitrifi- defined corresponding to NDEF > 10 lM. In the figures, NO2À sec- cation criteria, was applied to the WOA2005 (World Ocean Atlas, ondary subsurface peaks have been delimited arbitrarily by an iso-

2005) data. line corresponding to about half of the NO2À maximum (NO2maxÀ )to

be coherent with the NO2maxÀ intensity of each area. This NO2À crite- 2.1. CRIO (CRIterion on O2) criterion for OMZ estimation rion is in agreement with the conditions used previously for the eastern Paciﬁc Ocean and the northern Indian Ocean (e.g., Codispoti

The CRIO OMZ criterion is based on a characteristic O2 profile et al., 2001). defined for the Chilean OMZ from 200 data collected at 18 sta- tions during four cruises (2000–2002; between 20°S and 30°S), 2.3. WOA2005 database used for OMZ and NMZ estimations with high vertical resolution (5–10 m) sampling and an achieved accuracy of 0.5–1.0 lM(Paulmier et al., 2006). CRIO has been CRIO and denitrification criteria were applied using WOA2005 defined to take into account the OMZ core, but also the upper (World Ocean Atlas 2005) data obtained between 1893 and and lower O2 gradients, corresponding to the boundary layers 2004; this is the most recent and updated global O2 and nutrient between the core and the surrounding well oxygenated ocean, at database. From WOA2005 data, including WOCE (World Ocean the top and at the bottom of the OMZ, and contributing to the N Circulation Experiment) data and respecting WOCE quality stan- perturbation (e.g., Anderson et al., 1982). These three OMZ layers, dards, a yearly climatology (Boyer et al., 2006) for O2,NO3À and 2 covering an O2 continuum from aerobic to anaerobic conditions, PO4À global distributions was obtained and mainly used here. exhibit anomalies that differentiate them from the surrounding O2 climatology has been developed based on data from 632,888 more oxygenated seawater. profiles of bottle samples mainly taken in the last 30 years (>80% of

Since, from our biogeochemical point of view, OMZs should nec- the data; Boyer et al., 2006). The distribution of these O2 profiles is, essarily allow denitrification, an OMZ core (called CORE) has been a priori, correctly covering all the already identified hypoxic areas defined by O2 <20lM. Indeed, O2 <20lM corresponds to the (ENP, ESP, AS, BB; Kamykowski and Zentara, 1990). O2 accuracy maximum O2 concentration for which water-column denitrifica- and reproducibility are <10 lM and 2–10 lM, respectively. 2 tion was observed in situ (Smethie, 1987). The O2 <20lM concen- NO3À and PO4À WOA2005 climatology’s have been used to eval- tration also corresponds to a usual suboxic condition used to uate NDEF and is based on the same order of profile numbers and separate the aerobic (O2-respiration) from the denitrifying (NO3À- during the same periods as for O2, though 2.7 (233,125) and 1.6 respiration) activity (e.g., Oguz et al., 2000). In addition, this (400,399) times less abundant, respectively. Accuracy and A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 117

1 reproducibility are: 2 and 0.2 lM for NO3À; 0.03 and 0.02 lM CORE (O2 <20 lM: in red ). They are found mainly in the eastern 2 for PO4À. No global NO2À database is available with analytical or Paciﬁc (EP), and in the northern Indian (NI) semi-enclosed by the data-processing errors (Kamykowski and Zentara, 1991). NO2À has continents. In the EP, OMZs extend between 37°S and 52°N and therefore been used only for the four regions (ESP off Chile and from the coast out to 180°W (>10,000 km offshore); in the NI, Peru; ETNP – Eastern Tropical North Paciﬁc; AS) for which ade- OMZs extend from 23°N at the coast to 7°N (>1500 km south- quate NO2À data are available, from Nitrop-85 (C2–C3–C4), VERTEX wards) and including the area between 55°E and 100°E. In the I and II and Kanya cruises, respectively. Atlantic, LOZs have been detected in the North- and South-eastern

The calculation of the vertical and horizontal boundaries of the Atlantic Ocean (shades of yellow). These LOZs, with O2 P 20 lM OMZ–NMZ is based on WOA2005 climatology in net common data (>40 lM in the North; >20 lM in the South, except in very format (NetCDF), corresponding to a 1 1 zonal and meridional restricted areas over the continental margin (O reaching Â 2 interpolation and, temporally, to each season (Collier and Durack, 17–18 lM), are much less intense than the OMZs in the Pacific 2006). About three times more data are available near the coast and Indian Oceans, and are not considered in the present study. and in the northern hemisphere than in the middle of the oceans In the EP and NI, OMZs with distinct characteristics can be identi- and in the southern hemisphere, and that resolution induces an er- fied: ENP in the North, and ESP in the South; AS in the West, and BB ror of ±10% for the horizontal extent. Vertically, a linear interpola- in the East. In the EP, ENP and ESP, OMZs are in contact horizon- tion was performed between the standard depths of the O2 and tally, but neatly dissociated by three branches extending to the the nutrient concentrations (i.e. 10, 20, 30, 50 m, near the surface, West: ESP (0–37°S); ETNP (0–25°N); ESTNP (eastern subtropical and 1400, 1500, 1750, 2000 m). This low resolution, especially at North Pacific; 25–52°N), the highest latitude for a permanent the OXY, could induce an error up to 25% of the vertical three layers OMZ. There is no formation of OMZ at high latitude in the southern of an OMZ. Note also that it has only been possible to consider the hemisphere, except in the ESP off Chile, but only up to 37°S. Verti- upper and lower boundaries, but not the lateral boundaries because cally, these three branches can be clearly differentiated by their po- of the insufficiently high horizontal resolution of the WOA2005 sition. A deepening of 150 m from the ESP (CORE centered at a data: therefore, in the present study, the three OMZ layers have depth of 365 m, with upper CORE isobaths contouring the OMZ the same extent, horizontally. All OMZ dimensions and gradients at 400 m depth) to the ETNP (with upper CORE isobaths at were performed systematically using the same Ferret calculation 600 m), and of 430 m from the ETNP (530 m) to the ESTNP (with script for: (i) horizontal extent, following the depths of minimum upper CORE isobaths at 1200 m). Note the dissymmetry between

O2 concentration (the ‘‘Iso-minimum of O2” or ‘‘isominox”), be- the North and the South Paciﬁc, due to a closed coastline to the cause, as we will be seen (cf. CORE thickness, Table 1), the lowest West-North for the ENP compared to an open coastline oriented

O2 concentrations can be observed at different depths. For the South-East from the equator for the ESP. AS and BB are clearly dis- NMZ, the script calculates the NDEF maximum (‘‘isomaxndef”), by sociated horizontally by the southern end of India (and Sri Lanka), analogy with the ‘‘isominox”; (ii) vertical thickness, by sequential although with comparable isobaths around 200 m depth. Most of single integration over standard discrete sampling depths for each the OMZs are located in tropical latitudes (<25°); the ESP off Chile layer (OXY, CORE, LOG) and for the total OMZ layer (similar calcu- and the ESTNP OMZs being the only two located at higher latitudes. lations for NMZ); (iii) OXY and LOG O2 gradients by averaging from Regionally, each OMZ presents some spatial variability not de- the OXY and LOG O2 gradient determined at each standard discrete tailed here, as the two meridional AS components (Naqvi et al., sampling depth by differences between each consecutive depth: 2006): intense in the East (<8 lM), associated with the Indian con- i.e., (O O )/dz averaged over the whole thickness of the tinental margin; less intense in the central-western open ocean 2 z +dz À 2 z OXY and of the LOG. (>8 lM). In this study, the ESP OMZ is the only OMZ discussed In addition to the WOA2005 climatology, profiles and vertical mainly with respect to its three components (near the equator; sections from biogeochemical cruises have been used directly, with off Peru; off Chile), and has been especially illustrated off Peru the same or better accuracy (up to <1 lM for O2), to illustrate the and Chile. comparisons between: (i) different OMZs with representative sta- tions from the WOA2005 database (cf. Fig. 2): at 21°S, 71°W off 3.1.2. Vertical structure of the OMZs Chile; at 9°S, 85°W off Peru; at 12°N, 100°W in the ETNP; at The vertical OMZs structure is here described for each OMZ 15°N, 64°E in the AS; and at 15°N, 90°E in the BB; (ii) different cri- layer. OMZs present differences in O2 concentration (for the CORE), thickness and depth range, and O2 gradient (for the OXY and the teria (CRIO, NDEF, NO2À; cf. Figs. 4 and 5) from process study data: S4BGC at 21°S, 71°W off Chile (Paulmier, 2005), and N7 at 15°N, LOG). 64°E in the AS (US JGOFS); WOCE P21E at 17°S off Peru, and The COREs of ESP near the equator and in the BB are the least P19C at 90°W in the ETNP. intense (O2 P 7 lM), with O2 values four times higher than in all other OMZ COREs. The difference in minimal O2 values in CORE 3. Results is especially marked between: the BB (10 lM) and the AS (2 lM); the ESP near the equator (7 lM) and off Peru (3 lM); The results presented concern the extents of the OMZs and the and, in respect of the mean CORE O2 concentrations, between the NMZs. ESTNP (18 lM) and the ETNP (14 lM). In addition to the CORE intensity, the ESP OMZ, off Chile and Peru, has a mean O2 content 3.1. Extent of OMZs for OXY (89 and 81 lM, respectively) and for LOG (130 and 113 lM, respectively) that is 20% higher than for the other OMZs. The CRIO criterion including CORE (O2 <20 lM), OXY and LOG, The OMZs in the ETNP, the ESP off Peru and near the equator are applied to WOA2005 climatology, allows the identification the thickest and at the greatest depth OMZs (>3000 m; Fig. 2a). The (Fig. 1a), the evaluation of the vertical and horizontal structure OMZ off Chile is the thinnest (740 m) with the strongest OXY (Fig. 2 and Table 1) and the seasonality (Fig. 3 and Table 2) of all (2.1 lM/m). The CORE in AS OMZ is the thickest (>750 m). The the OMZs. The following Sections 3.1.1–3.1.3 always refer to Table 1. COREs off Chile and Peru and in the BB are the shallowest (from 160 m). The COREs in the ESP near the equator, the ETNP and the 3.1.1. Identification of OMZs This section always refers to Fig. 1a. The most intense OMZs in 1 For interpretation of color in Fig. 1, the reader is referred to the web version of the open ocean are identified by the suboxic conditions in their this article. 118

Table 1 Horizontal extent (Area), vertical characteristics (OXY; CORE; LOG) in terms of thickness and depth range, O2 concentration ([O2]) and O2 gradients for the main and most intense permanent OMZ regions in the open ocean, with maximal mixed and euphotic layers depth

OMZ regions Total OMZ Oxycline Core Lower O 2 gradient (LOG) Mixed Euphotic f 6 Layer layer m Area 10 Thickness m Thickness Mean [O 2] lM O2 Gradient Thickness Mean [O 2] lM Thickness Mean [O 2] lM O2 Gradient e 2 a b c d d c d c d d Depth km (%) [Zmin ;Zmax ] m (%) (max [O 2]) lM/m (max) m (%) (min [O 2]) m (%) (max [O 2]) lM/m (max) b m [Cmin ;Cmax ] Global ocean 30.4 ± 3 3360 ± 800 440 ± 170 65 ± 58 (202) 1.7 ± 0.6 340 ± 280 15 ± 3 (2) 2580 ± 870 100 ± 43 (140) +0.04 ± 0.06 –– À [10; 3370] (13%) ( 5.9) (10%) [450– (77%) (0.1) À 790] Eastern Eastern Chile: (18– 0.4 ± 0.1 740 ± 720 150 ± 50 89 ± 78 (238) 2.1 ± 0.8 160 ± 120 15 ± 2 (3) 430 ± 740 130 ± 38 (166) +0.03 ± 0.02 44 60–120 À Pacific South 37°S; [10; 750] (20%) ( 4.7) (22%) (58%) (0.1) 113–128 (2009) 80 Oceanography in Progress / Ruiz-Pino D. Paulmier, A. À (EP) Pacific 70–82°W) [160; 320] (ESP) Peru: (0–18 °S; 0.6 ± 0.1 3740 ± 1020 160 ± 40 81 ± 70 (214) 1.8 ± 0.6 340 ± 160 (9%) 13 ± 2 (3) 3240 ± 1010 113 ± 43 (150) +0.03 ± 0.02 69 75–90 À 80–90°W) [10; 3750] (4%) ( 4.7) [170; 510] (87%) (0.1) À Equatorial 4.7 ± 0.5 3360 ± 560 270 ± 100 65 ± 56 (201) 1.7 ± 0.6 190 ± 170 (6%) 15 ± 2 (7) 2900 ± 520 105 ± 38 (143) +0.03 ± 0.02 139 60–110 À component: [10; 3370] (8%) ( 3.5) [280; 470] (86%) (0.1) À (0–18°S; 75– 120°W) 5.7 ± 0.6 3490 ± 660 260 ± 100 66 ± 58 (238) 1.7 ± 0.6 190 ± 170 (5%) 15 ± 2 (3) 3040 ± 650 108 ± 39 (220) +0.03 ± 0.02 148 60–120 À (19%) [10; 3500] (7%) ( 4.7) [270; 460] (87%) (0.1) À Eastern Eastern 12.4 ± 1 3560 ± 760 310 ± 130 59 ± 55 (197) 1.8 ± 0.7 420 ± 290 14 ± 3 (2) 2830 ± 860 99 ± 43 (141) +0.04 ± 0.02 103 30–90 À North Tropical (41%) [10; 3570] (9%) ( 5.9) (12%) (79%) (0.1) À Pacific North Pacific [320; 740] (ENP) (ETNP): (0–25°N; 75– 180°W) Eastern Sub- 8.2 ± 1 2950 ± 760 830 ± 160 76 ± 64 (240) 1.2 ± 0.3 230 ± 130 (8%) 18 ± 1 (3) 1890 ± 810 100 ± 45 (141) +0.04 ± 0.02 175 – À Tropical North (27%) [20; 2970] (28%) ( 2.6) [850; 1080] (64%) (0.1) À Pacific (ESTNP): (25–52°N; 75–180°W) North Arabian Sea (AS): 2.5 ± 0.2 2980 ± 680 230 ± 160 63 ± 58 (191) 1.6 ± 0.4 760 ± 340 13 ± 4 (2) 1990 ± 770 93 ± 44 (127) +0.05 ± 0.03 96 150–200 À Indian (7–23°N; 55–77°E) (8%) [10; 2990] (8%) ( 3.3) (26%) (67%) (0.1) À (NI) [240; 1000] Bay of Bengal (BB): 1.6 ± 0.2 2400 ± 600 170 ± 30 69 ± 62 (191) 1.9 ± 0.5 310 ± 160 16 ± 2 (10) 1920 ± 670 85 ± 43 (126) +0.05 ± 0.02 81 – À (8–20°N; 80–100°E) (5%) [10; 2410] (7%) ( 3.3) (13%) [180; (80%) (0.1) À 490]

Values are calculated from annual and regional averages (WOA2005 database) using CRIO criterion for the deﬁnition of each layer. Note that for OXY and LOG Thickness, upper and lower limits correspond to [ Zmin :Cmin ] and [Cmax :Zmax ], respectively. ± Indicates: for Area columns, errorbar on the calculation mainly due to the horizontal resolution of the WOA2001 climatology; for other columns, classical standard deviation on the whole considered areas. a % of each OMZ area on the total OMZ area. b Zmin and Cmin , and Zmax and Cmax are the upper and lower depth of the total OMZ and the CORE, respectively. c % of the OXY or CORE or LOG thickness on the total OMZ thickness. d Maximal concentration (for OXY at Zmin and LOG at Zmax ) or absolute value of O 2 gradient (for OXY and LOG O2 gradient), and minimal concentration for CORE. e 3 Maximal depth from 0.03 kg m À density criterion ( De Boyer Montégut et al., 2004 ). f Range of the maximal depth for the euphotic layer (from JGOFS 2003 database; for ESP off Chile, Ulloa, Pers. Com; for ESTNP and BB, no data available found). A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 119

Oxygen (µM) Zm Ze 0 40 80 120 160 200 240 0 40 80 120 160 200 240 (m) 0 44 500 50 60 EAST PACIFIC 69 80 ESPC 1000 100 103 90 Chile 1500 Peru 150 ETNP 2000 200 2500 250

3000 300 b 3500 ETNP ESPP 350 c d 4000 a (m)

0 Depth 500 50 AS 81 NORTH INDIAN 100 1000 BB 96 1500 150 175 2000 BB 200 2500 250

3000 300 f 3500 AS 350 g h 4000 e

Fig. 2. O2 proﬁles for 0–2700 m (a and e) and for 0–300 m (b and f), mean maximal mixed-layer (Zm: c, g) and euphotic layer (Ze: d, h) for the main OMZs in the Paciﬁc (upper row, a and b: ESP off Chile (21°S; 71°W) and off Peru (9°S; 85°W); ETNP (12°N; 100°W)) and the Indian (lower row, e and f: AS (15°N; 64°E); BB (15°N; 90°E)) Oceans from WOA2005 data at representative locations for each OMZ. Horizontal striped segments indicate the lower OMZ limit (a and e). Horizontal lines represent the upper OMZ depth (b and f). For acronyms, see in Fig. 1a.

ESTNP are the deepest (from 280 m down to 850 m). The mean Despite differences, a common vertical structure can be seen for CORE thickness of each OMZ is between 4 (AS) and 20 (ESP) times all the OMZs: smaller than the thickness of the total vertical OMZ structure. The (i) OXY: strong ( 1.7 lM/m; P1.2 lM/m), four times more oxy- thickness of the CORE, but also of the OXY, relative to the total OMZ genated ( 65 lM) than CORE, shallow from a depth of 10–20 m, thickness is quite comparable ( 10%), except for the AS where the intercepted by euphotic zone and by the maximal annual mixed- CORE thickness is three times larger than that of the OXY and for layer depth; (ii) CORE: intense (1 6 O2 <20lM), highly O2-defi- the ESTNP where it is the contrary (1/3). The large thickness of cient ( 15 lM, reaching minima <10 lM), extending over several the ESTNP OXY induces the less strong OXY ( 1.2 lM/m). Gener- hundred meters (>300 m on average) between 160 and 1080 m À ally, LOGs in the EP OMZs are relatively thicker (>2800 m) and less depth; (iii) LOG: one order of magnitude thicker ( 2580 m), less strong with higher mean O2 concentrations (100–130 lM) than in strong ( 0.04 lM/m), and 35% more oxygenated ( 100 lM) than the NI (85–93 lM). in the OXY. The euphotic layer in the ETNP is the shallowest (from 30 ± 30 m deep) and in the AS (up to 200 ± 25 m), but with a mean 3.1.3. Horizontal extent of OMZs depth for all the OMZs intercepting the OXY (Euphotic Layer col- Horizontally, all OMZs extend (identified here by ‘‘isominox” umn; Z in Fig. 2d and h). Note however that the information on depth; Fig. 1a), over an area of 30.4 106 km2 (±10%), a significant e Â the euphotic layers is only indicative, on average, because of the surface accounting for 8% of the present global ocean surface. The high spatial and temporal variability of this parameter (up to a fac- ETNP OMZ, covering 12.4 106 km2 (41% of the entire OMZs’ sur- Â tor of 2). The maximal mixed-layer depths off Chile and Peru, and face), is the largest (Table 1), followed by the ESTNP (27%) and the in the BB and the AS are the shallowest (44–96 m); and for the ESP (19%) OMZs. The ENP OMZ (ETNP+ESTNP) covers 20.6 Â ESTNP, the ESP near the equator and the ETNP, the deepest (103– 106 km2, about 78% of the total Pacific OMZs. This difference in ex-

175 m; Zm in Fig. 2c and g). But, the maximal mixed-layer depths tent between the ENP and the ESP OMZs has probably to be associ- for all the OMZs intercept the OXY, hence allow the OMZ come ated with the so-called conveyor belt which is more impoveriched partially in contact with the ocean–atmosphere interface during in O2 in the North than in the South Paciﬁc. In the ESP OMZ, the the year. Because the relationship between factors derived from components near the equator covering a surface of 4.7 106 km2 Â climatology ignores the probable temporal disconnect between and representing about 82% of the total ESP OMZ, appear more statistical tendencies, comparisons between Z Z and OMZs important than the better known ESP off Peru and Chile. The small- e À m should be considered with much caution. est OMZs are those found in the Indian Ocean, in the AS and the BB 120 A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128

Fig. 3. O2 distribution (lM; at ±2 lM intervals) at depth where O2 concentration is minimal, according to the WOA2005 climatology and to the CRIO criterion (O2 <20lMin the CORE) applied in spring (a), summer (b), fall (c) and winter (d). For OMZs acronyms, see in Fig. 1a.

(respectively 8% and 5% of the global OMZ surface). However, both and the O2 concentration (Mean [O2] 15 ± 1 lM; Min[- areas are only 30% lower than that of ESP OMZ. O2] 0 lM). Among the seasonally permanent OMZs, the ETNP Considering the EP and NI OMZs as permanent open-ocean and ESP OMZs, as at the global scale, do not show significant sea- OMZs (without taking into account the low seasonality: see here sonal changes, although the extreme western end could detach below), with a vertical extent of 3360 m (±800 m), OMZs represent from the rest of the OMZ in: summer for the ETNP (at 185°W; overall a mean volume estimated at 102 ± 15 106 km3 (7 ± 1% of Fig. 3b); fall for the ESP (at 130°W; Fig. 3c). Â the ocean volume). The CORE (O2 <20lM) of all OMZs taken to- However, for the other permanent OMZs, a seasonal variability gether occupies a volume of 10.3 106 km3, about one tenth of has been detected, at least for one of the structural components, Â that occupied by the all the OMZs as a whole, including OXY and horizontally or vertically or in intensity, and here illustrated the LOG. In volume, the biggest OMZ COREs are in the ETNP and the most completely by all the structural components of the AS OMZ. AS, due to their large horizontal extent (representing almost half During the spring–summer transition, the AS OMZ CORE presents the total OMZ extents in the global ocean) and vertical extent a thickening of 20% (from 640 m to 790 m), associated with a (up to twice the mean thickness of the CORE), respectively. shoaling from 280 to 220 m of the upper CORE limit. This thickening is associated with a CORE intensification (>30%, from 16 lMto 3.1.4. Seasonality of OMZs 12 lM for the mean O2 concentrations), largely confirmed by the The climatological seasonal variability of the whole OMZ has minimal O2 concentrations decreasing from 1 lM to 0.1 lM. Hori- been analyzed for the total OMZ horizontal extent, their CORE zontally, on the contrary, the AS OMZ presents a contraction of thickness and vertical depth range, and their O concentration in 10% (from 3 to 2.7 106 km2). During the summer–fall–winter– 2 Â the CORE. This variability is commented for the five OMZs already spring transition, the variability is opposite. The AS OMZ CORE con- described annually, and then for two new OMZs appearing season- tracts vertically by 20% (from 790 to 720 m, then 710 m, then ally. This whole section refers only to Fig. 3 and Table 2, unless 640 m, for the four mentioned seasons, respectively), with a deep- otherwise indicated, and to boreal seasons. ening of the OMZ CORE upper limit by 60 m (from 220 m to 280 m Fig. 3 shows at the global scale that the five main OMZs, already depth). The CORE becomes less intense by 30%, with O concen- 2 described from Fig. 1a (ESTNP, ETNP, ESP, AS, BB), are always pres- trations increasing from 12 lMto13lM, then to 16 lM. Horizon- ent and with comparable horizontal extents (32.2 < 34.6 tally, the OMZ extends by 10%, from 2.7 to 3 106 km2. Â Â 106 km2) and features for each of the four seasons. Therefore, sea- This seasonal tendency, observed for the AS and especially for sonality would hardly affect the OMZs horizontal extent ( 10% be- the marked spring–summer transition, can also be observed for tween 0.1 and 3.0 106 km2), the CORE thickness and vertical the ESTNP, the ESP and the BB. But the seasonal changes are ob- Â depth range (330–340 m with C 440–470 m; cf. note b, Table 2), served only in some of the structural components, and only in min A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 121

some seasonal transitions. Indeed, between spring and summer, the ESTNP OMZ contracts horizontally by 7% (from 9.2 to 8.6 106 km2). The ESP OMZ off Chile thickens by 67% (from

– Â 120 m to 200 m) with an intensiﬁcation of its CORE by >15% (from

. 15 lM to 13 lM). Between fall and winter, the OMZ CORE in the ESP off Peru contracts vertically by nearly 20% (from 360 m to

Table 1 300 m), with a deepening of the CORE upper limit of 20 m (from c – ]) 160 m to 180 m). Between spring and summer, the OMZ CORE in 2 the ESP at the equator intensiﬁes by 7% from 16 lMto15lM for the mean O2 concentrations, and from 8 lMto5lM for the

M (min[O minimal O2 concentrations. Finally, always between spring and l

] summer, the BB OMZ contracts horizontally by 10% (from 1.7 to 2 1.55 106 km2). Â In addition to the permanent OMZs, an OMZs map for each season allows the detection of two new OMZs that disappear seasonally, and which therefore cannot be observed on an annual map Winter Spring Summer Fall 18 ± 1 (15) – Core mean [O (Fig. 1a):

(i) in the West Bering Sea (WBS: 45–65°N; 175–210°W) appearing mainly in winter (Fig. 3d; 2.2 106 km2), and Â more or less disappearing in the other seasons (Fig. 3a–c;

– <0.3 106 km2). This WBS OMZ reaches 7% of the total Â OMZs’ surface (2.2 106 km2), which is comparable to the Â AS OMZ surface (2.5 106 km2: Total OMZ Area in Table Â 1), and with a deep (between 690 and 1100 m) and less

intense (mean O2 of 17–18 lM) CORE;

Summer Fall – (ii) in the Gulf of Alaska (GA: 52–65°N; 120–175°W) appearing in fall–winter–spring (Fig. 3c, d, a; >0.4 106 km2), with a Â marked presence in fall and spring (1.1 and 0.7 106 km2, Â respectively) and an attenuated presence in winter (0.4 106 km2). GA more or less disappears in summer b ) Â 6 2 (Fig. 3b; 60.1 10 km ). Therefore, the GA OMZ should min Spring

C Â ( present, as the other permanent OMZs generally do, a hori-

m zontal contraction between spring and summer by a factor of 7 (from 0.7 to 0.1 106 km2), whereas its CORE thickens, Â up to a factor of >5 (from 40 to 210 m). Note that the GA ]) for the main OMZ region in the open ocean

2 OMZ is in contact with the ESTNP mainly in the fall, as well

Core thickness as in winter and spring (Fig. 3a, c, d), probably corresponding to a northern prolongation of the ESTNP associated with a

connection and an O2 deﬁcit transfer between both OMZs. In fact, according to the CORE vertical depth range between 1.1 ± 0.1 160 ± 150 (670) 40 ± 100 (770) 210 ± 160 (850) 180 ± 230 (840) 19 ± 1 (13) 18 ± 1 (10) 18 ± 1 (14) 18 ± 1 (8)

concentration ([O 670 and 1060 m depth, with a CORE upper limit (e.g., Cmin 2

a = 840 m in spring), GA and ESTNP are at comparable

l depths, and present less intense COREs (Core Mean [O2]of 18–19 lM). Taking into account the WBS (in winter) and the GA (in fall and spring), the total OMZs surface increases 2 up to 10% rising to 33.5 106 km2. km

6 Â Thus, in the OMZs in which a significant seasonality could be observed, there is a marked spring–summer transition associated with, on average: (i) CORE thickening of 4.4%, generally associated with a CORE upper limit shoaling (between 10 and 30 m ). 34.3 ± 3 34.6 ± 3 32.2 ± 3 32.8 ± 3 330 ± 280 (440) 340 ± 280 (470) 340 ± 300 (450) 340 ± 280 (440) 15 ± 1 (0) 15 ± 1 (0) 15 ± 1 (0) 15 ± 1 (0) Total OMZ area 10 Winter Spring Summer Fall Winter depth); (ii) a horizontal OMZ contraction by 1.3%; (iii) CORE intensification associated with a diminution in the mean O2 con- Table 1 centration up to 4 lM, as in the AS; (iv) OXY and LOG with higher ESPPESPEqESP 0.6 ± 0.1 4.9 ±ETNP 0.5ESTNP 0.6 ± 0.1 12.6 6.0 4.7 ± ± ± 1 0.6 0.5AS 8.3 ± 1 0.6BB ± 0.1 6.0 5.0 13.6 ± ± ± 0.6 0.5 1 2.7 9.2 ± ± 0.2 0.6 6.1 1 ± ± 12.7 0.1 0.6 4.4 ± 1.6 ± 1 ± 0.5 0.2 300 3.0 ± ± 190 0.2 190 (180) ± 8.6 170 5.6 1.7 ± ± (270) ± 1 0.6 0.2 13.5 ± 350 1 ± 2.7 150 ± 180 0.2 (150) 190 ± ± 170 170 (290) 1.55 (260) ± 0.2 340 410 ± ± 160 8.0 300 170 ± (180) 180 (320) ± 1 2.7 ± 160 ± 170 (280) 0.2 (280) 1.7 ± 360 0.2 440 ± ± 170 300 190 (160) 710 170 ± (330) 200 ± ± 180 ± 350 170 130 (270) 370 (230) (270) ± (870) 170 13 (170) 430 ± ± 1 310 14 (0.8) 640 190 ± (320) 240 ± ± 1 ± 340 180 130 (5) 270 (280) (250) ± (890) 180 13 (200) 400 ± ± 1 280 (2) 790 14 (320) 220 ± ± ± 350 1 130 16 200 (220) (0.8) ± ± (850) 1 160 (8) (200) 14 ± 1 12 720 15 (0.2) 250 ± ± ± ± 1 400 1 140 380 (0.5) (240) (2) ± (840) 190 15 (160) 13 ± ± 1 1 13 13 (5) (0) 17 ± ± ± 1 1 1 15 (1) (0) 15 ± (0.4) ± 1 1 (2) (0.5) 15 14 17 ± ± ± 1 1 1 16 (5) 15 (0) (0.7) ± ± 1 1 16 (1) (0.9) ± 1 (3) 17 ± 1 14 (4) ± 1 12 (1) ± 1 17 (0.1) ± 1 (3) 17 ± 1 13 (2) ± 1 (0.3) 15 ± 1 (1) GA 0.4 ± 0.1 0.7 ± 0.1 0.1 ± 0.1 O2 concentration less intense, associated with an oxygenation increase up to 9 lM and 6 lM, respectively, as in the BB OMZ (not shown in Table 2). Especially during the fall–winter transition, concentration; often below the significativity ±2

2 the OMZs generally present a more gradual seasonality and opposite to the spring–summer transition. These observations of vertical thickening and CORE intensiﬁca- tion, observed in spring–summer, are in agreement with an in- ] = minimal O

2 crease of hypoxia (less oxygenation) reported during the same = CORE upper limit depth (Cf. period by Kamykowski and Zentara (1990). This seasonality also min

GA OMZ considered as disappearedC due to the uncertainty. min[O matches the local observations of OMZ formation over the shelf c a b Global ocean Permanent OMZ for all seasons ESPC 0.3 ± 0.1 0.2 ± 0.1OMZ 0.2 with ± seasonal 0.1 disaparition WBS 0.2 ± 0.1 2.2 140 ± ± 0.2 110 (180) 120 0.2 ± ± 110 0.2 (160) 200 ± 100 0.3 ± (150) 0.2 170 ± 120 (150) 0.2 ± 0.2 15 ± 1 (5) 170 ± 160 (690) 15 – ± 1 (3) 13 ± 1 (1) 15 ± 1 (0.9) Values are calculated from seasonal and regional averages (WOA2005 database) using the CRIO criterion for the CORE deﬁnition. For OMZs acronyms, cf. the end of the text. For errorbars (±), cf. note of Table 2 Seasonal OMZ changes in horizontal extent (Area),OMZ in regions CORE thickness and mean O and/or shoaling in spring–summer, associated with a stronger 122 A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128

upwelling (e.g., off Chile; Graco, 2002). But here, the OMZs season- d ality appears more complex. In fact, vertical thickening is associated with horizontal contraction during spring–summer, and CORE intensification with more oxygenated OXY and LOG; and in- versely in fall–winter. The anti-correlated seasonal variability in M (max [NDEF]) the intensity of the CORE and the OXY–LOG oxygenation suggest l probable spread of the O2 deficit between the CORE and its boundary layers. This proposed seasonality cannot be detailed further here, mainly because of a lack of data to identify significant tendencies for each season.

3.2. Extent of the denitriﬁcation (NMZs) zone 13 ± 3 (39) Mean [NDEF] in NMZ 12 ± 1 (29) 15 ± 2 (27) 14 ± 2 (24) 12 ± 1 (15) 12 ± 2 (27) 12 ± 1 (22) <10 (<10) 12 ± 1 (20) 11 ± 1 (20)

This whole section refers always to Table 3, unless otherwise ) indicated. Table 1 3.2.1. Identification e 0 Total OMZ (From 3360 740 3740 3360 3490 3560 2950 2980 Denitrification regions (NMZs) in the open ocean have been 2400 identified on a global scale using the ‘NDEF > 10 lM’ criterion ) (Fig. 1b). As for OMZs, NMZs are found in the EP, between 37°S and 25°N and from the coast out to 160°W (>8000 km offshore), and more confined in the NI, from 23°N at the coast to 8°N Table 1

(>1400 km southwards) and including the area between 55° and e 0 OMZ Core (From 340 160 340 190 190 420 90°E only. But a NMZ is also found in the Arctic Ocean, mainly asso- 230 ciated with the large western continental shelves, as in the Chukchi and Beaufort seas, in agreement with the N* analysis by Gruber and Sarmiento (1997). Except the Arctic region (high latitude: >60°N), [80; 240] [230; 460] where there is no water-column OMZ (Fig. 1a), four main NMZs [220; 480] [50; 270] [140; 250] can be distinguished in the ETNP, ESP, AS and BB OMZs, but not [160; 220] in the ESTNP. Regionally, the ETNP NMZ is much more conﬁned to the c coast (out to 140 W) than the OMZ, and presents a dissociated lo- ] m

° b max

cal NMZ far offshore near the equator and 140°W. The highest Zn NDEFs (up to 22 lM) are conﬁned close to the coast, whereas the ; min

rest of the ETNP presents NDEFs mainly between 11 and 18 M. Zn Thickness [ 90 ± 110 [70; 160] 230 ± 320 (68% and160 7%) ± 80 (100% and 22%) 220 ± 70 (65% and 6%) 60 ± 50 (32% and110 2%) ± 90 (58% and 3%) 260 ± 240 (62% and 7%) 0 (0% and 0%) 400 ± 270 (53% and 13%) [190; 590] 760 l 90 ± 140 (29% and 4%) [130; 220] 310 The ESP NMZ is also more restricted to the south (between 10°S and 37°S). The NMZ off Peru is the most intense (with higher NDEF M layer. up to 24 lM) close to the coast, where the rest of the ESP presents l NDEFs mainly between 11 and 18 lM. The AS NMZ covers quite well the OMZ extent with a weak extension restricted to the north

of 10°N and with homogenous NDEFs of 12 lM (highest values e OMZ (this study) NMZ T (%) 0 >19 lM in the northeastern part close to the coast). Contrary to the AS, the BB NMZ extent is very restricted close to the east coast of India between 80°E and 90°E, with a low mean NDEF of 2 km a

11 lM. 6 . 3.2.2. Vertical extent of the denitrification (NMZs) zones NMZ (%) Area 10 15.0 ± 1 (49%) 30.4 ± 3 0.4 ± 0.1 (100%) 0.4 ± 0.1 0.6 ± 0.1 (100%)4.2 ± 0.6 0.4 ± 0.1 (67%) 5.7 ± 0.6 7.8 ± 1 (56%) 12.4 ± 1 0 ± 0.1 (0%) 8.2 ± 1 2.3 ± 0.2 (80%) 2.5 ± 0.2 0.7 ± 0.1 (37%) 1.6 ± 0.2 4.7 ± 0.5 Vertically, the AS and ETNP NMZs are characterized as the deepest (lower limit between 480 and 590 m depth) and thickest (be- Table 1 M Maximum Zone. The last column indicates the NDEF concentration, to allow an estimation of the NMZ intensity. Values are calculated from the annual and regional tween 260 and 400 m depth). The ESP NMZ is the shallowest l (upper limit from 50 m; lower limit from 220 m) and one of the thinnest ( 230 m), with the highest NDEF (up to 27 lM off Chile and 24 lM off Peru). The BB NMZ is also one of the shallowest (upper limit at 130 m) and the thinnest (90 m), with the lowest NDEF for an NMZ (11 M). There is no NMZ in the ESTNP Chile: Peru: Equatorial component: 3.2Eastern ± 0.3 Tropical (68%) North Pacific (ETNP): 4.7 ± 0.5 l Eastern Sub-Tropical North Pacific (ESTNP): (NDEF < 10 ± 2.5 lM). The Arctic NMZ without a corresponding OMZ, is the shallowest (from 70 m depth), the thinnest (90 m, as for the BB) and one of the most intense (13 nM up to 39 lM) NMZs. The vertical mean NMZ structure (Table 3) is confirmed by the

illustrating profiles (off Chile and in the AS; Fig. 4) and sections (off are the upper and lower depth, and NMZ T the Thickness of the NDEF > 10 Eastern South Pacific (ESP) Eastern North Pacific (ENP) Arabian Sea (AS): Bay of Bengal (BB): À Peru and in the ETNP; Fig. 5) of NDEF and NO2 . Regarding the typ- max ical profile off Chile, the NMZ (Fig. 4b) has a thickness of 145 m be- Zn N): ° tween 150 and 295 m depth, in agreement with the NO2À maximum and

(Fig. 4c). For the AS, the NMZ (Fig. 4e) has a thickness of 370 m be- min % of the NDEFmax on% OMZ of for the the NDEFmax area. onZn OMZ CORE and totalMaximal OMZ NDEF for concentration. the thickness,Arctic respectively. Ocean is not corresponding to a water-column OMZ. Paciﬁc (EP) (NI) tween 230 and 600 m depth, that appears largely thicker than sug- (67–90 c a e b d averages (WOA2005 database). For errorbars (±), cf. note of Table 3 Comparison between NMZ and OMZ from theRegions CRIO criterion of this study, for the horizontal extent (Area) and vertical (Thickness) characteristics of the main permanent OMZ regions in the open ocean NMZ corresponds to NDEF >10 (±2.5 for the errorbars) Global ocean Eastern North Indian gested by the NO2À maximum (Fig. 4f). The thickness of the NMZ off Arctic Ocean A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 123

Fig. 4. Proﬁles of O2 (a and d), NDEF (b and e: NDEF>10 lM shaded) and NO2À (c and f: NO2À > NO2maxÀ =2 lM shaded) at representative locations (21°S; 71°W) off Chile (top row) and (15°N; 64°E) in the AS (bottom row). In a and d, backgrounds correspond to vertical extent of the OMZ (dashed area) and of the CORE (grey area). A typical mean vertical O2 proﬁle from WOA data ‘outside OMZs’ (classical O2 minimum) is shown in pointed line.

Fig. 5. Vertical sections of O2, NDEF and NO2À (a–c: Peru at 17°S; d–f: ETNP at 90°W). Contours (thick black lines) correspond to: O2 =20±2lM (a,d); NDEF = 10 ± 2.5 lM (b,e); NOÀ NOÀ /2 lM (c and f). 2 ¼ 2max 124 A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128

Peru (Fig. 5b) and in the ETNP (Fig. 5e) are slightly larger than sug- OMZ CORE extends southwards out to 3000 km from the coast: gested by the NO2À maximum (Fig. 5c and f). Consequently, for all i.e., almost three times farther than for the NMZ (Fig. 5d and e). the NMZs, the vertical thickness appears generally higher than sug- On a global scale, the NMZ evaluation presents the same ten- gested by the NO2À maximum. dency to underestimate the OMZ CORE extent than was previously noted for each OMZ. Horizontally, OMZs (Fig. 1a) correspond to 3.2.3. Horizontal extent of the denitrification (NMZs) zones NMZs (Fig. 1b) in the ETNP, ESP, AS and BB, but the extent of the Horizontally (Fig. 1b; Table 3), all NMZs, without taking into ac- NMZs is more restricted than that of the OMZs out only to count the Arctic NMZ, extend over an area of 15.0 106 km2 (±7%), 2000 km offshore in the EP, and only 1000 km southwards from Â i.e., 4% of the ocean surface. The ETNP NMZ, covering 7.8 the coast in the NI. The ETNP NMZ is about 2500 km closer to the Â 106 km2 (52% of the entire NMZs’ surface), is the largest, followed coast than the OMZ, and the ESP NMZ is limited to the southern by the ESP NMZ (28%). The NMZ in the northern hemisphere covers part, between 10° and 40°S, about 1000 km less extended towards about 65% of the total Pacific NMZs. The smallest NMZs are those the equator than the OMZ. The global NMZs area extent is found in the Indian Ocean: the AS (15%), and especially the BB 15 106 km2, two times less (49%) than that of the overall OMZs Â (5% of the global NMZs’ surface). The Arctic NMZ is not taken into (see note a, Table 3). For each OMZ, the same tendency to the under- account here, because it does not have a corresponding water-col- estimation of the OMZ by the NMZ was observed between 37% in umn OMZ, although it covers a significant area of 5.4 106 km2 the BB and 100% in the ESP off Chile and Peru. Vertically, all the Â (one-third of all the other NMZs). OMZ COREs are associated with denitrification detected by the

NMZ extent can be compared qualitatively to the NO2À maxi- presence of an NMZ (NDEF P 11 lM), but underestimated by mum extent (Figs. 4 and 5). Off Peru, the NMZ extends westwards NMZ by a factor of 1.5, between 0% in the ESP off Chile (100% of 1600 km from the coast, i.e., much farther than the NOÀ maxi- the CORE) and 68% in the ESP near the equator (32% of the CORE). 6 2 mum (Fig. 5b and c). Likewise, for the ETNP, the NMZ extends In synthesis, for all OMZs (Tables 1 and 3), the NMZ underesti- southwards 1000 km from the coast, i.e., almost as far as for mates the OMZ extent by the following factors: (i) horizontally, be- 6 the NOÀ maximum (Fig. 5e–f). The NOÀ maximum, although with tween 1 and 3 (weighted average: 2); (ii) vertically, between 1 2 2 a smaller extent, confirms the NMZ extent and the presence of and 7 (weighted average: 1.5) for the CORE and 15 for the total denitrification. OMZ including OXY and LOG. In volume, OMZs (7% of the oceanic Total NMZs’ volume, considering a vertical extent of 230 m, volume) should be underestimated, by the denitrification criteria, represents a mean volume of 3.45 ± 0.05 106 km3 ( 0.03 ± by a factor of 3 for the CORE and of 30 for the total thickness. Â 0.001% of the ocean volume). In volume, the biggest NMZs are in the ETNP (58% of the total NMZ volume) and the AS (27%), as for 4.2. Comparisons between OMZs–NMZs and previous evaluations the OMZs. The smallest NMZ volume is in the BB, 2% of the total NMZ. The NMZ seasonality could not be studied at all, because of The OMZs’ structure is relatively different from those of the insufficient availability of data for each season. ‘‘classical O2 minima”, found in all the oceans and between 500 m and 2500 m depth (mean average depth about 1500 m;

Wyrtki, 1962). All OMZs attaining the lowest CORE O2 concentra- 4. Discussion tion of 2 lM are >50 times more intense than classical O2 minima, which are characterized by O2 concentrations of 50 and 200 lM OMZs and NMZs characterized in this study were compared be- for the Paciﬁc and Atlantic Oceans, respectively (WOA2005 data- tween themselves, and with the classical O2 minimum and previ- base). OMZs, which may be located from 10 m depth, are 50 times ous evaluations. Then, the choice of a criterion to take into shallower and with a vertical extent over hundreds of meters, account the entire OMZ volume associated with OMZs is discussed. whereas even if the classical O2 minima are present in all oceans, the most intense OMZs are restricted to the EP and NI Oceans.

4.1. Comparison between OMZs and NMZs extents These differences between OMZs and classical O2 minima are shown for the ESP off Chile and for the AS (Fig. 4a and d). The To characterize the differences between the volume of OMZs ESP OMZ (Fig. 4a) CORE off Chile is between 160 and 320 m depth, and denitrification zones, the extents of the OMZs and NMZs were with O2 reaching <1 lM, whereas the classical O2 minimum is be- compared. These comparisons are first illustrated with profiles off tween 1200 and 1600 m depth and with a minimum O2 concentra- Chile and in the AS (Fig. 4) and with vertical sections off Peru and in tion of about 40 lM. The situation off Chile is unique, presenting the ETNP (at 17°S and 90°W; Fig. 5), then confirmed on a global two differentiated O2 minima, one corresponding to the OMZ and scale. the other to the classical O2 minima. The classical O2 minimum Locally, and vertically, on a typical OMZ profile off Chile, the off Chile, although five times less intense ( 100 lM), may corre- CORE and NMZ are located between 160 and 320 m depth, and be- spond, by hemispheric symmetry in the EP, to the deep ESTNP tween 150 and 295 m depth, respectively (Fig. 4a and b). The NMZ OMZ (CORE up to 1080 m). On the AS OMZ (Fig. 4d), CORE is lo- would thus underestimate the OMZ CORE thickness (160 m) by cated between 240 and 1100 m depth, with O2 also reaching the about 9 ± 3%. For the AS, the CORE and NMZ are located between detection limit (<1 lM), whereas the classical O2 minimum is at 240 and 1000 m depth, and between 230 and 600 m depth, respec- 2000 m depth, with a O2 minimum of 50 lM. tively (Fig. 4d–e). The vertical NMZ extent ( 370 m) is more than For OMZs (Fig. 1a, Table 1), the main documented extent, based two times smaller than the CORE extent. Off Peru, the CORE and on O2 data all focused in the last decades, was estimated by differ- NMZ are located on average between 170 and 510 m depth, and ent criteria for the O2 concentration: at the global scale for the between 120 and 350 m depth, respectively (Fig. 5a–b). The NDEF tropical Pacific (ESP, ETNP), by Karstensen et al. (2008); from local denitrification criterion therefore underestimates the OMZ CORE studies on the vertical extent by Anderson et al. (1982) for the thickness (340 m) by 30% (110 m). For the ETNP, the CORE thick- ETNP, ESP off Peru, and for the AS, by Morrison et al. (1999). ness is 420 m, between 320 and 740 m depth, but the NMZ is Horizontally, OMZ CRIO evaluation, based on the suboxic condi-

15 ± 4% thinner and 1.3 times shallower (less than 250 m depth; tion O2 <20lM, corresponds qualitatively to the same regions iden- Fig. 5d–e). Horizontally, the area of the OMZ CORE off Peru area tified by the hypoxia (O2 <8lM) map (Kamykowski and Zentara, extends westwards out to 1800 km from the coast, which is 10% 1990), except for the offshore ESP near the equator and ESTNP com- farther than for the NMZ (Fig. 5a–b). Likewise, for the ETNP, the ponents. This relatively global agreement suggests that the CORE of A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 125 all the OMZs approaches hypoxia or at least O2 6 10 lM(Table 1). CORE. But the presence of denitrification did not allow the delim- Moreover, with WOA2005 climatology and CRIO, the ESTNP and itation of the entire volume associated with the OMZ perturbation. ETNP OMZs are clearly dissociated, because OMZ extents have been Two main reasons may be given: (i) denitrification does not exactly derived from the ‘‘isominox” calculation following the O2 minimum correspond to the CORE volume, which is intrinsically related to in the OMZ CORE. This dissociation between ESTNP and ETNP is in the denitrification process; (ii) denitrification is probably not the agreement with the frequency diagram of the number of hypoxia only important biogeochemical process associated with the OMZs. observations (Kamykowski and Zentara, 1990), showing dissocia- The first difficulty is due to the determination of different vol- tion between the tropical (between 0° and 30°N) and subtropical umes for the canonical (i.e. classically anaerobic) denitrification.

(between 30° and 52°N) north latitudes. The OMZ CRIO evaluation The O2 threshold value, beneath which denitrification occurs, is is also in agreement with the description of Karstensen et al. variable, from <0.8 lM in the eastern tropical Pacific (Goering, (2008), especially in the eastern tropical Pacific. Note that OMZs ex- 1968) to <20 lM in an ENP fjord (Smethie, 1987), and seems to de- tend far into the open ocean (>8000 km for ESTNP and, closer to the pend on the region and on each local denitrifying bacteria commu- equator, >10,000 km for ETNP); that is 30 times farther offshore nity. Thus, the same criterion cannot be applied to determine the than the extent of the OMZs in contact with the seafloor denitrification volume for all the OMZs. In addition, the determina- (1148 106 km2; Helly and Levin, 2004). Although their extent tion of denitrification is often performed using different criteria Â would always cover the coastal seafloor, the OMZ extent in the open and at least three parameters (NDEF, NO2À peak, O2 conditions), ocean is not correlated with the sedimentary extent of each OMZ. which do not describe the same volume (see Figs. 4 and 5).

Vertically, in all previous work on the O2 deficit, only OMZ CORE The second difficulty is due to OMZ CORE underestimation by has been considered, but OXY and LOG have not been taken into canonical denitrification criteria (cf. Section 4.1). Denitrification account. OMZ CORE thickness is globally in agreement with the occurs at O2 concentrations that are always less than 20 lM estimates (O2 < 4.5 lM) of Karstensen et al. (2008) for the eastern (Smethie, 1987). Thus, the canonical denitrification criteria could tropical Pacific (ESP, ETNP). Locally, the OMZ CORE thicknesses are underestimate the OMZ CORE volume (O2 <20lM) only, and, a in agreement within ±7% for the ESP off Peru (340 ± 160 m), but fortiori, the total OMZ volume including OXY and LOG. In addition, only within 30 ± 7% for the ETNP (420 m), which is higher than the basic explanation of the OMZ CORE underestimation by the the previous estimates (Anderson et al., 1982) evaluated from a NMZs extent is that denitrification and hence NMZ detection is twice as constraining O2 criterion (<10 lM) than CRIO. For the mainly controlled by the low level of O2 in the OMZ CORE, to which AS, the thickness (760 ± 340 m) is in agreement with that of a pre- denitrifying bacteria would be very sensitive (Codispoti et al., vious study on the whole basin (O2 < 4.5 lM; Morrison et al., 2001). Indeed, the ESP (especially off Chile and Peru) and the ETNP, 1999), although using an O2 criterion even more constraining, by which form intense NMZs (22 6 NDEFmax 6 27 lM), correspond to a factor of 4, than CRIO. The highest difference in thickness is then the most intense OMZ CORE (2 6 O2 min 6 3 lM). The BB NMZ, the mainly attributable to the studies based on local cruises, hence not most restricted compared to the other OMZs and the least intense representative of the whole OMZ basins as defined herein. NMZ (mean NDEF 11 lM), is associated with the least intense In volume, the global OMZs CORE is about seven times larger OMZ CORE (O2 always >10 lM). In the same way, the ESTNP than the usually considered suboxia volume (0.1% of the oceanic OMZ, which does not have an NMZ (NDEF always <10 lM), corre- volume) proposed in Codispoti et al. (2001). For each oceanic basin, sponds to the less intense (Mean [O2]=18lM) and deeper (below the OMZs CORE appears 16, 1000 and 120 times larger than 850 m) OMZ CORE and the thickest OXY. Therefore, the absence of that of the OMZs in the North (0.44 106 km3) and South an NMZ (ESTNP) or a low NDEF (BB) could be mainly due to a 6 3 Â 6 3 (0.001 10 km ) Pacific, and Indian Ocean (0.002 10 km ), esti- slightly more oxygenated OMZ (>2 lMofO2 compared with the Â Â mated by Karstensen et al. (2008) with the more constraining O2 other OMZs), but also to a lower level of bacterial activity because < 4.5 lmol/kg condition, respectively. of a lesser supply of organic matter, as well as because of the cold For NMZs (Fig. 1b; Table 3), volume comparison with the results and different environmental conditions. The ESTNP is the deepest of previous denitrification extent studies reported in the literature (sinking organic matter is largely already degraded) and the cold- is not possible, because of the lack of information in the vertical est OMZ (3 < 5 °C instead of >15 °C for the other OMZs). The BB is dimension for a whole OMZ basin. At the global scale, our NMZs highly subject to the continental influence of two of the world’s surface is about two times larger than the evaluation from Hattori biggest rivers (Brahmaputra, Ganges), which impose hydrological (1983), mainly because the BB and the ESP near the equator and off and biogeochemical conditions (e.g. temperature, salinity, organic Chile were not considered. Locally, denitrification comparisons matter quality, bacteria composition) that are very different from were made only for the ETNP, the ESP off Peru, and the AS the marine ones. (Fig. 1a; Table 3). The ETNP NMZ (7.8 106 km2: Table 3) is about On the other hand, an ‘‘OMZ border-specific denitrification” Â 14% more extensive, and by factors 2.5 and 7, compared with the can appear at the OXY of all OMZs with O2 >20lM, and could ex- denitrification surface evaluated by Goering et al. (1973), Cline plain why the NMZ would be slightly shallower than the OMZ and Richards (1972) and Codispoti and Richards (1976), respec- CORE (cf. Figs. 4 and 5). This specific denitrification, off Peru, tively. The NMZ off Peru (0.6 106 km2) is about 1.8 times less was suggested by Codispoti and Christensen (1985), supported extensive than the denitrification surface evaluated by Codispoti by dN15 analysis in the AS (Naqvi et al., 1998), and recently and Packard (1980). The AS NMZ (2.3 106 km2)is 46 times named OLAND (oxygen-limited autotrophic nitrification–denitrifi- Â more extensive than the denitrification surface evaluated by Deus- cation: non-canonical denitrification from NO2À produced by cou- er et al. (1978). These comparisons do not only document the way pled nitrification; Brandes et al., 2007). Moreover, it is known that in which the presence of denitrification is determined, but also the denitrification can proceed in the presence of O2 (Zehr and Ward, spatial documentation and sampling strategy; previous studies 2002), outside the CORE boundaries, at a thin OXY but also at a based on cruises were generally more localized than the horizontal thick LOG (Li et al., 2006). This hypothesis could be confirmed estimate made here for the whole OMZ basin. by the potential existence of an oxygenated denitrification near

O2 saturation (>200 lM; Patureau et al., 1994). In this case, deni- 4.3. Why denitrification criteria could not delimitate the entire OMZ? trification criteria could overestimate OMZ CORE volume, or neglect to take into account the volume in which non-canonical The principal studies and evaluations in OMZs are focused on denitrification (as OXY and LOG) or other biogeochemical the limited volume in which denitrification occurs, mainly in the processes could occur. 126 A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128

The second reason why denitriﬁcation could not delimitate the focus the attention for future studies. Otherwise, the OMZ and entire OMZ is probably due to the existence of other biogeochem- NMZ reference states proposed in the present study could help to ical processes related to the N cycle, which could locally affect evaluate past and future variations in OMZs extent, and to under-

NDEF and NO2À in the O2 minimum. Consequently, from indirect stand the biogeochemical cycles and biological anomalies induced indices as NDEF, it is difficult to capture, just the denitrification by the OMZ formation. process. In OXY and/or LOG, nitrification, remineralization, OLAND (cf. above), anammox could appear (e.g., Smethie, 1987; Ward 5. Conclusion et al., 1989; Paulmier et al., 2006; Thamdrup et al., 2006; Brandes et al., 2007). Photosynthesis could also occur near the OXY, because The global ocean area and volume occupied by the most intense 2 of the availability of light, as suggested by the secondary fluores- OMZs (O2 <20lM) have been evaluated: 30.4 ± 3 millions of km cence peak associated with Prochlorococcus and Synechococcus and 102 ± 15 millions of km3, accounting for, respectively, 8% and communities (e.g., Liu et al., 1998). Among the biogeochemical 7% of the global ocean. These results suggest that the extent of processes, nitrification can be extremely active above (OXY) and the OMZs was previously underestimated in the open ocean. below (LOG) the OMZ CORE (Anderson et al., 1982), and coexis- Horizontally, this study allowed to distinguish the permanent tence between denitrification and an aerobic process is possible OMZs, despite seasonality changes (contraction; expansion) of (e.g., nitrification; Farias et al., 2007). More complex mechanisms 10–15%, and the seasonal OMZs which completely disappear for could also affect the nitrogen cycle, such as the recently discovered several months during the year. The permanent OMZs are found anammox reaction (NHþ NOÀ N under anaerobic conditions; in the same four regions as those in which hypoxia and denitrifica- 4 þ 2 ! 2 Kuypers et al., 2003). Consequently, a denitrification criterion does tion have been previously identified: ETNP, ESP, AS, BB. But, the not allow the determination of the volume associated with all the present study also points out the formation of permanent OMZs biogeochemical processes occurring in the entire three OMZ layers. in subtropical latitudes, such as the deep ESTNP, between 25°N NDEF may also be intensified by benthic denitrification, which and 52°N, and very well dissociated from the well known ETNP. could be more intense than in the water column, as observed in OMZs also formed seasonally in high latitudes: WBS (45–65°N; the eastern AS OMZ on the continental margin (Naqvi et al., 175–210°W) with a surface similar to that of the AS OMZ, which 2000). The NMZ detected in the Arctic Ocean was found mainly appears mainly in winter, and in the GA (52–65°N; 120–175°W) on the large western continental margins, whereas no Arctic which only disappears in summer. The seasonality of the OMZs OMZ has been detected in the water column. This would be asso- appears more complex, with a vertical expansion often associated ciated with a high level of denitrification in the suboxic–anoxic with an opposite horizontal contraction and CORE intensification sediments (e.g., in the Bering, Chukchi and Beaufort Seas; Devol during spring-summer, and is difficult to investigate from the cur- et al., 1997), which transmits a high NDEF signal detected here in rently available global data. the water column (as for NO2À; Kamykowski and Zentara, 1991). Vertically, the OMZs, which are overall little oxygenated The Arctic NMZ would also be associated with a horizontal advec- ( 88 lM), show common characteristics including three layers: tion from the North Pacific (Yamamoto-Kawai et al., 2006). This (i) an OXY: strong ( 1.6 lM/m), four times more oxygenated implies both that: (i) a less intense OMZ with O2 >20lM, as the ( 65 lM) than the CORE, shallow (from 10–20 m depth) interfac- LOZs in the eastern Atlantic, could be formed without water- ing with the euphotic zone, and with the maximal annual mixed column canonical denitrification; (ii) an NMZ could be formed layer depth; (ii) a CORE: intense (O2 <20lM), highly deficient in without intense (O2 <20lM) water-column OMZ, as in the Arctic. O2 ( 15 lM), between 160 and 1080 m depth; (iii) a LOG: one Due to the potential co-existence of different OMZ-specific pro- order of magnitude thicker ( 2580 m), less strong ( 0.04 lM/m), cesses, the use of the same criterion, directly based on O2 distribu- and 35% more oxygenated ( 100 lM) than for the OXY. The OMZs, tion, as CRIO, could be better adapted to consider the influence of localized in the EP, NI, WB and GA, and reaching O2 concentrations the whole OMZ and its specific biogeochemical processes. In par- down to <1 lM in the CORE in the subsurface, are different from ticular, CRIO avoids the underestimation of the biogeochemical the relatively well-known ‘‘classical O minimum”, which is 50 2 activities linked to nitrogen cycle but also to carbon remineraliza- times more oxygenated and found in the intermediate waters tion and to O2 consumption, at the OXY and LOG. Based on a sub- (1000–1500 m) of all the oceans. The CORE of all OMZs occupies oxic criterion for the CORE, CRIO generally corresponds to the a volume ten times smaller than the one occupied by the whole various definitions of the OMZ: oxygen-minimum layer (OML; OMZ, including OXY and LOG. OXY presents a similar thickness Goering, 1968), oxygen-deficient zone or layer (ODZ or ODL; Naqvi to that of the CORE, 10% of the total OMZ thickness. LOG is one et al., 2006). In addition, the common characteristics of all OMZs order of magnitude thicker (80%) than the CORE. This study pro- with respect to OXY, CORE and LOG all validate the CRIO criterion, posed therefore a characterization of the vertical OMZs structure which was chosen from the characteristics of the OMZ off Chile. with three layers, instead of considering the lowest O2 concentra- However, in the future, and depending on the improvements in tions in the CORE only, as previously performed. the detection of O2, CRIO could be adapted for specific and regional All OMZs are associated with NMZs (defined as NDEF > 10 lM), studies and processes in the OMZs (e.g., canonical denitrification; except the ESTNP OMZ and the LOZs in the eastern Atlantic. The sulfate-reduction and sulphide release). Moreover, oxygen mea- global NMZ here evaluated at 15 ± 1 millions km2 without includ- surements and CRIO do not account for microzone anoxia that ing the Arctic NMZ, suggests that the global denitrification zone might be associated with particles like marine snow in the low- was underestimated at least by a factor of 2 in the previous studies. oxygen water column. Horizontally, because the global extent of the NMZs is three times The similarities among all the OMZs, characterized from CRIO, smaller than that of the OMZs, denitrification zones could also allow us to hypothesize that all OMZs are the consequences of sim- underestimate the OMZs area. Vertically, the NMZ estimate leads ilar dynamical (low ventilation) and biogeochemical (O consump- to an underestimation of the extent by factors of 1.5 for the CORE 2 tion) processes. However, the marked differences between OMZs and 15 considering the total OMZ tickness. The reasons why and the classical O2 minima suggest that, regionally, either differ- denitrification zones underestimate the OMZ extents are: (i) the ent mechanisms or different intensity control each contraction denitrification occurs mainly in a restricted volume compared to and extent of an OMZ. The maintaining of the OMZ formed by three the entire OMZ CORE volume (O2 <20lM); (ii) other biogeochem- layers (OXY, CORE, LOG), related to a large range of O2 concentra- ical OMZ-specific processes, especially linked to the nitrogen cycle, tions, involves complex biogeochemical processes which should are not taken into account by the use of indirect indices A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128 127

(e.g., NDEF) for the detection of the denitriﬁcation. Consequently, Foreign Affairs) and the University of Paris VI. We thank V. Garçon the extents of the NMZs and OMZs suggest that the ocean could for critical reading of an early manuscript version, C. Duarte, H.J. lose more nitrogen than previously thought. Minas and M. Graco for the successful discussions. Thanks to R. The most extended NMZs in the ETNP and ESP represent 52% Grifﬁths for correcting the English, and C. Provost, L. Mortier, M. and 28% of the global NMZ area, respectively, and the NMZ in the Crépon, MN Houssais and JM André, for helping and encouraging AS is the thickest (760 m). The ETNP and the ESP (off Chile and this work.

Peru) form the most intense NMZs (NDEFmax between 22 and 27 lM), also corresponding to the most intense OMZ CORE (O2 References min between 2 and 3 lM). For the seasonal WB and GA OMZs, there are not enough data to evaluate their associated NMZ. On the other Anderson, J.J., Okubo, A., Robbins, A.S., Richards, A.F., 1982. A model for nitrite and hand, a polar NMZ was detected in the Arctic Ocean mainly on the nitrate distributions in oceanic oxygen minimum zones. Deep-Sea Research 29, large western continental margins associated with high denitrifica- 1113–1140. tion in suboxic-anoxic sediments, whereas no Arctic OMZ was de- Arrigo, K.R., 2005. Marine microorganisms and global nutrient cycles. Nature 437, 349–355. tected in the water column. NMZ extents are always lower than Bange, H.W., Rapsomankis, S., Andrae, M.O., 1996. Nitrous oxide in coastal waters. those of the OMZs, and the differences could be related to the avail- Global Biogeochemical Cycles 10, 197–207. ability of O : the more intense the OMZ is, the smaller is the differ- Béthoux, J.P., 1989. Oxygen consumption, new production, vertical advection and 2 environmental evolution in the Mediterranean Sea. Deep-Sea Research 36 (5), ence between the dimensions of the NMZ and of the OMZ. Thus, 769–781. the ESTNP, BB and ESP near the equator, which are the most oxy- Boyer, T.P., Antonov, J.I., Garcia, H.E., Johnson, D.R., Locarnini, R.A., Mishonov, A.V., genated OMZs, have the most underestimated OMZs in volume Pitcher, M.T., Baranova, O.K., Smolyar, I.V., 2006. World Ocean Database 2005. Levitus, S. (Ed.), NOAA Atlas NESDIS 60, Washington, DC, 190pp., DVDs. (<50%), by the criterion of denitrification. Brandes, J.A., Devol, A.H., Deutsch, C., 2007. New developments in the marine The OMZ extent and vertical structure, determined in the pres- nitrogen. Cycle Chemistry Review 107, 577–589. ent study, could be used to evaluate past and future variations of Broecker, W.S., Peng, T.H., 1982. Tracers in the sea, Eldigio Press. Cannariato, K., Kennett, J., 1999. Climatically related millenial-scale fluctuations in OMZs, linked to both climatic (temperature increase) and environ- strength of California margin oxygen-minimum zone during the past 60 ky.. mental (fertilization by nutrients) changes. To improve the estima- Geology 27, 975–978. tion of the extent of the OMZ, it will be necessary to reduce Chan, F., Barth, J., Lubchenko, J., Kirincich, A., Weels, H., Perterson, W., Menge, A., uncertainties in vertical extent (about 10%) for all the OMZs, 2008. Emergence of Anoxia in the California current large marine ecosystem. Science 319, 920. including the less known ESTNP OMZ. We recommend systematic Cicerone, R., Oremland, R.S., 1988. Biogeochemical aspects of atmospheric methane. sampling programs adapted to ultra-low O2 (<1 lM) for the CORE, Global Biogeochemical Cycles 2 (655-2), 661. with high resolution: vertically, 5–10 m or continuous (as with Cline, J.D., Richards, F.A., 1972. Oxygen-deficient conditions and nitrate reduction in the eastern tropical North Pacific. Limnology and Oceanography 17, 885–900. new microsensors); horizontally, <1/4°. These recommendations Codispoti, L.A., Christensen, J.P., 1985. Nitrification, denitrification, and nitrous should be taken into account in the new O2-measurements pro- oxide cycling in the eastern tropical southeast Pacific Ocean. Marine Chemistry jects (e.g., ARGO-Oxygen, recently proposed by Gruber et al., 16, 277–300. Codispoti, L.A., Packard, T.T., 1980. Denitrification rates in the eastern tropical South 2007; the European project Oxywatch), especially in the most Pacific. Journal of Marine Research 38, 453–477. extensive and intense OMZs such as the ETNP. The first step to- Codispoti, L.A., Richards, F.A., 1976. An analysis of the horizontal regime of wards new OMZ studies is to: (i) documentate the intensity, extent denitrification in the eastern tropical North Pacific. Limnology and Oceanography 21, 379–388. and seasonally changes in all the OMZs; (ii) elucidate the still un- Codispoti, L.A., Brandes, J.A., Christensen, J.P., Devol, A.H., Naqvi, S.W.A., Paerl, known physical and biogeochemical mechanisms that form and H.W., Yoshinari, T., 2001. The oceanic fixed nitrogen and nitrous oxide maintain intense and extensive OMZs in the modern ocean. Other- budgets: moving targets as we enter the anthropocene? Science Marine 65 (2), 85–105. wise, all our efforts to predict OMZ extent and impact on ecosys- Collier, M.A., Durack, P.J., 2006. The CSIRO netCDF version of the NODEC World tems and biodiversity, using mathematical 3-D coupled Ocean Atlas. CSIRO Marine and Atmospheric Research Paper 2005 015, 48. circulation and biogeochemical models, as a consequence of De Boyer Montégut, C., Madec, G., Fischer, A.S., Lazar, A., Iudicone, D., 2004. Mixed anthropogenic fertilization and climate change, will be in vain. layer depth over the global ocean: an examination of profile data and a profile- based climatology. Journal of Geophysical Research 109, 1–20. Deuser, W.G., Ross, E.H., Mlodzinska, Z.J., 1978. Evidence for and rate of 6. Main acronyms denitrification in the Arabian Sea. Deep-Sea Research 25, 431–445. Devol, A.H., Codispoti, L.A., Christensen, J.P., 1997. Summer and winter denitrification rates in western Arctic shelf sediments. Continental Shelf AS: Arabian Sea; BB: Bay of Bengal; BlS: Black Sea; BS: Baltic Sea; Research 17, 1029–1050. C: Carbon; CORE: OMZ core (O2 <20lM); CRIO: criterion on O2 to Dugdale, R., Goering, J., Barber, R., Smith, R., Packard, T., 1977. Denitrification and characterize the OMZ OXY, CORE and LOG; ENP/ESP: eastern hydrogen sulfide in Peru upwelling region during 1976. Deep-Sea Research 24, 601–608. North/South Pacific; ESPC/ESPEq/ESPP: eastern South Pacific off Falkowski, P.G., 1997. Evolution of the nitrogen cycle and its influence on the Chile/near the equator/off Peru; EP: eastern Pacific; ESTNP: eastern biological sequestration of CO2 in the ocean. Nature 387, 272–275. subtropical North Pacific; ETNP: eastern tropical North Pacific; GA: Farias, L., Paulmier, A., Gallegos, M., 2007. Nitrous oxide and N-nutrient cycling along the oxygen minimum zone off northern Chile. Deep-Sea Research 34, Gulf of Alaska; ‘‘isominox”/”isomaxndef”: depth of minimal O2 164–180. concentration (‘‘iso-mininimum of oxygen”)/maximal NDEF; LOG: Fernández-Alamo, M.A., Färber-Lorda, J., 2006. Zooplankton and the oceanography of the eastern tropical Pacific: a review. Progress in Oceanography 69, 318–359. lower O2 gradient; LOZ: low O2 zone (e.g., in the eastern Atlantic); Goering, J.J., 1968. Denitrification in the oxygen minimum layer of the eastern N: Nitrogen; NI: northern Indian Ocean; NDEF: nitrate deficit tropical Pacific Ocean. Deep-Sea Research 15, 151–164. (NO3À); NMZ: NDEF (>10 lM) maximum zone; OMZ: oxygen mini- Goering, J., Richards, F.A., Codispoti, L., Dugdale, R., 1973. Nitrogen fixation and mum zone; OXY: oxycline; PG; Persian Gulf; RS: Red Sea; SWACM: denitrification in the ocean: biogeochemical budgets. In: E. Ingerson (Ed.), southwest African continental margin; WBS: western Bering Sea; Proceedings of the International Symposium on Hydrogeochemistry and Biogeochemistry, Ontario, pp. 12–27. WOA2005: World Ocean Atlas 2005; Ze/Zm: euphotic layer/maximal Graco, M., 2002. Variación temporal de la producción de amonio en un area de mixed-layer depth. surgencias de Chile Central (36°S). Factores y procesos involucrados en su reciclaje. PhD Thesis. Gray, J.S., Wu, R.S., Or, Y.Y., 2002. Effect of hypoxia and organic enrichment on the Acknowledgments coastal marine environment. Marine Ecology Progress Series 238, 249–279. Gruber, N., 2004. The dynamics of the marine nitrogen cycle and its influence on This study was supported by a French CNRS Ph.D. Fellowship to atmospheric CO2. The Ocean Carbon Cycle and Climate, Follows and Oguz, Dordrecht, 97–148. A. Paulmier and financial support was provided by IRONAGE (Euro- Gruber, N., Sarmiento, J.L., 1997. Global patterns of marine nitrogen fixation and pean Union Program), the ECOS Sur program (French Ministry of denitrification. Global Biogeochemical Cycles 11, 235–266. 128 A. Paulmier, D. Ruiz-Pino / Progress in Oceanography 80 (2009) 113–128

Gruber, N., Doney, S.C., Emerson, S.R., Gilbert, D., Kobayashi, T., Körtzinger, A., Naqvi, S.W.A., Naik, H., Pratihary, A., D’Souza, W., Narvekar, P.V., Jayakumar, D.A., Johnson, G.C., Johnson, F.S., Riser, S.C., Ulloa, O., 2007. The ARGO–oxygen Devol, A.H., Yoshinari, T., Saino, T., 2006. Coastal versus open-ocean program. White paper from the Argo Steerring Comité, 60. denitrification in the Arabian Sea. Biogeoscience Discussions 3, 665–695. Hattori, A., 1983. Denitrification and dissimulatory nitrate – reduction. In: Nitrogen Oguz, T., Ducklow, H.W., Malanotte-Rizzoli, P., 2000. Modeling the distinct vertical in the Marine Environment. Academic Press, New York, pp. 455–497. biogeochemical structure of the Black Sea: dynamical coupling of the oxic, Helly, J.J., Levin, L.A., 2004. Global distribution of naturally occurring marine suboxic, and anoxic layers. Global Biogeochemical Cycles 14 (4), 1331–1352. hypoxia on continental margins. Deep-Sea Research 51, 1159–1168. Patureau, D., Davison, J., Bernet, N., Moletta, R., 1994. Denitrification under various Joos, F., Plattner, G., Stocker, T., Körtzinger, A., Wallace, D., 2003. Trends in Marine aeration conditions in Comamonas sp., strain SGLY2. FEMS Microbiology Ecology Dissolved Oxygen: Implications for Ocean Circulation Changes and the Carbon 14, 71–78. Budget. EOS 21, 197–204. Paulmier, A., 2005. Zones de Minimum d’Oxygène (OMZs) de l’océan moderne. Une Kamykowski, D., Zentara, S.-J., 1990. Hypoxia in the world ocean as recorded in the étude focalisée dans le Pacifique Sud Est Ph.D. Thesis, Université Curie, Paris, historical data set. Deep-Sea Research 37, 1861–1874. France, 248pp. Kamykowski, D., Zentara, S.J., 1991. Spatio-temporal and process-oriented views of Paulmier, A., Ruiz-Pino, D., Garçon, V., Farias, L., 2006. Maintaining of the East South nitrite in the world ocean as recorded in the historical data set. Deep-Sea pacific oxygen minimum zone (OMZ) off Chile. Geophysical Research Letter 33, Research 38, 445–464. L20601. Karstensen, J., Stramma, L., Visbeck, M., 2008. Oxygen minimum zones in the Richards, F.A., 1965. Anoxic basins and fjords. In: Ripley, J.P., Skirrow, G. (Eds.), eastern tropical Atlantic and Pacific oceans. Progress in Oceanography 77, 331– Chemical Oceanography. Academic Press, London and New York, pp. 611–645. 350. Rogers, A.D., 2000. The role of the oceanic minima in generating biodiversity in the Kiene, R.P., Bates, T.S., 1990. Biological removal of dimethylsulphide from seawater. deep sea. Deep-Sea Research II 47, 119–148. Nature 345, 702–705. Smethie, W.M., 1987. Nutrient regeneration and denitrification in low oxygen Kuypers, M., Sliekers, A., Lavik, Schmidt, M., Jørgensen, B., Kuenen, J., Siminghe fjords. Deep-Sea Research 34, 983–1006. Damste, J., Strous, J., Jetten, M., 2003. Anaerobic ammonium oxidation by Stramma, L., Johnson, G.c., Sprintall, J., Mohrholz, V., 2008. Expanding oxygen – anammox bacteria in the Black Sea. Nature 422, 608–611. minimum zones in the tropical oceans. Science 320, 655–658. Levin, L.A., 2002. Deep-Ocean Life Where Oxygen Is Scarce. American Scientist. Thamdrup, B., Dalsgaard, T., Jensen, M.M., Ulloa, O., Farıas, L., Escribano, R., 2006. September–October. Anaerobic ammonium oxidation in the oxygen-deficient waters off northern Li, Y-H., Menviel, L., Peng, T-H., 2006. Nitrate deficits by nitrification and Chile. Limnology and Oceanography 55, 2145–2156. denitrification processes in the Indian Ocean. Deep-Sea Research 53, 94–110. Tyrrell, T., 1999. The relative influences of nitrogen and phosphorus on oceanic Liu, H., Campbell, L., Landry, M., Nolla, H., Brown, S., Constantinou, J., 1998. primary production. Nature 400, 525–531. Procholorococcus and Synechococcus growth rates and contributions to Ward, B.B., Glover, H.E., Lipschultz, F., 1989. Chemoautotrophic activity and production in the Arabian Sea during the 1995 Southwest and Northeast nitrification in the oxygen minimum zone off Peru. Deep-Sea Research 36, Monsoons. Deep-Sea Research 45, 2327–2352. 1031–1051. Morrison, J.M., Codispoti, L.A., Smith, S.L., Wishner, K., Flagg, C., Gardner, W.D., Wyrtki, K., 1962. The oxygen minima in relation to ocean circulation. Deep-Sea Gaurin, S., Naqvi, S.W.A., Manghnani, V., Prosperie, L., Gundersen, J.S., 1999. The Research 9, 11–23. oxygen minimum zone in the Arabian Sea during 1995. Deep-Sea Research 46, Yakusev, E.V., Neretin, L.N., 1997. One-dimensional modelling of nitrogen and 1903–1931. sulphur cycles of the aphotic zone of the Black and Arabian Seas. Global Naqvi, S.W.A., Yoshinari, T., Jayakumar, D.A., Altabet, M.A., Narvekar, P.V., Devol, Biogeochemical Cycles 11, 401–414. A.H., Brandes, J.A., Codispoti, L.A., 1998. Budgetary and biogeochemical Yamamoto-Kawai, M., Carmack, E., McLaughlin, F., 2006. Nitrogen balance and implications of N2O isotope signatures in the Arabian Sea. Nature 394, 462– Arctic throughflow. Nature, 443. 464. Zehr, J.P., Ward, B.B., 2002. Nitrogen cycling in the ocean: new perspectives on Naqvi, S.W.A., Jayakumar, A., Narvekar, P.V., Naik, H., Sarma, V., D’Souza, W., Joseph, processes and paradigms. Applied Environmental Microbiology 68, 1015–1024. S., George, M.D., 2000. Increased marine production of N2O due to intensifying Zumft, W.G., 1997. Cell biology and molecular basis of denitrification. Microbiology anoxia on the Indian continental shelf. Nature 408, 346–349. and Molecular Biology Reviews 61, 533–616.